On Bayesian analysis of the proportional hazards model